|Publication number||US5701501 A|
|Application number||US 08/023,693|
|Publication date||Dec 23, 1997|
|Filing date||Feb 26, 1993|
|Priority date||Feb 26, 1993|
|Publication number||023693, 08023693, US 5701501 A, US 5701501A, US-A-5701501, US5701501 A, US5701501A|
|Inventors||Jayanti L. Gandhi|
|Original Assignee||Intel Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (7), Non-Patent Citations (4), Referenced by (17), Classifications (15), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
______________________________________Loop: Set bit Control bit branch if clear status bit______________________________________
1. Field of the Invention
The present invention relates to a method and apparatus for executing atomic instructions. More particularly, the present invention relates to a method and apparatus for insuring the atomicity of atomic instructions.
2. Art Background
An atomic instruction is a computer instruction which requires completion of the instruction prior to execution of subsequent instructions. The atomicity of the instruction may be required, for example, because the result of the execution of the atomic instruction is required for execution of the subsequent instructions, or because the atomic instruction requires a maintenance of the state at the time the instruction is initiated. Currently, an atomic operation may be implemented using long routines in microcode or by stalling the pipeline processor until execution of the atomic operation is completed. These approaches have fundamental drawbacks of being inflexible when the need for the addition of a new atomic operation arises because the microcode must be modified or major hardware changes need to be made to accommodate the new atomic operations each time. Furthermore, the microcode change will also be accompanied by the addition of a new instruction and this will require modifications to a well established software tools chain.
In the method and apparatus of the present invention, a simple but effective architecture provides for a fast and flexible software mechanism for executing an atomic operation which is transparent to processor hardware changes. This architecture is adaptable to future performance enhancements and works with varying pipeline stages.
The present system provides for a simple, but effective, method to ensure the atomicity of instructions using existing hardware. Two bits from a visible register, a control bit and a status bit, are utilized. The control bit is read and write accessible. The status bit is read only, but is updated with the data of the control bit a predetermined number of clock cycles after the control bit is written to. To prevent subsequent instructions from executing prior to the atomic instruction's completion, the control bit is set and the status bit is monitored. Once the status bit is set, subsequent instructions are executed. The delay required between the update of the control bit and the status bit ensures that the atomic instruction completes execution prior to execution of subsequent instructions. Furthermore, as the status bit is read only and is updated only by the control bit, the state of the status bit cannot be modified. Therefore, the atomicity of the instructions are preserved without the performance overhead incurred in prior art systems.
The objects, features and advantages of the present invention will become apparent to one skilled in the art in view of the following detailed description in which:
FIG. 1 is a simple block diagram of an exemplary computer system.
FIG. 2a is a block diagram illustrating the architecture for implementing an atomic operation and FIG. 2b is illustrative of a register which is visible to the user and contains a control and status bit.
FIG. 3a and 3b are examples of code which implements and atomic operation in accordance with the architecture of the present invention.
In the following description for purposes of explanation, numerous details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that these specific details are not required in order to practice the invention. In other instances, well known electrical structures and circuits are shown in block diagram form in order not to obscure the present invention unnecessarily.
An illustrative computer architecture for implementing an atomic operation is embodied in a computer system such as that shown in FIG. 1. The processor 10 communicates via a bus 20 to a plurality of peripheral devices such as memory 30, input/output devices 40 and other systems and buses 50. Preferably, the architecture for implementing an atomic operation is included in the processor 10 and is controlled by instructions issued and received through the peripheral devices 30, 40 and 50. Exemplary systems and processors are those manufactured by Intel Corporation, Santa Clara, Calif.
A simplified exemplary architecture is shown in FIG. 2a. A first buffer 60 outputs a first status bit referred to as a control status bit which is visible to the user as a bit in a register, referred to herein as "Register A" and illustrated by FIG. 2b. This bit is read and write accessible. The buffer 60 has write line input 65, a read input 70, data in 80 and data out 85. The bit can be manipulated using the read, write and data inputs which, for example, can be controlled by the user reading and writing the control bit visible in Register A. The output 85 of buffer 60, which is the control bit, is further input to a second buffer 90, which subsequently outputs a status bit visible at a user visible register. However, the status bit visible is read only. The status bit is updated only by internal hardware which inputs the control bit, output by buffer 60, to the buffer 90 after a predetermined delay, as controlled by the delay 95. Thus, the state of the status bit, although visible and readable by the user, is not controllable except through the setting of the control bit.
To execute an atomic instruction, the code employing the atomic instruction takes advantage of this structure to ensure the atomicity of the instruction. Examples of instructions are shown in FIGS. 3a and 3b. FIG. 3a can be used, for example, to disable interrupts. As there is latency between the issuance of a disable interrupt instruction and the subsequent execution of the disabling of the interrupt, it is important that this instruction is atomic. To ensure this, the following loop can be instituted. The predetermined bit of a predetermined register is identified as the control bit. For example, in an Intel 960 CA processor, bit 31 of the IMSK register (SF1) may be used. The setting of this bit also initiates the disable instruction to interrupts. The next instruction is a branch if clear (BBC) instruction which checks the status bit to determine if the bit is set or reset. If the bit is clear or reset, the processor branches to the target location identified in the instruction, in this example, Loop 1. In accordance with the architecture shown in FIG. 2a, the set bit command causes a write operation to bit C of the first buffer 60. The output of the buffer 60 is also input to delay 95, which delays the signal for the predetermined amount of time and then inputs the value to the second buffer 90 for the status bit. Thus, referring back to FIG. 3a, a branch if clear instruction referencing the status bit will cause the processor to return to Loop 1 and cycle down through to the branch if clear statement until the status bit is set to 1. Once the status bit is set to 1, the processor proceeds to the next statement containing the next instruction to be executed.
It is apparent that the use of the BBC statement is illustrative and other branch instructions can be used. The resultant effect of the code is that the instructions subsequent to the atomic instruction are delayed a specified period of time to permit the execution of the atomic instruction, thereby insuring the atomicity of the instruction. The amount of time is tailored to the completion of the atomic instruction execution (i.e., associated hardware function such as disabling interrupts) and can be done in a few clock cycles as opposed to the 40 minimum clock cycles found in the prior art. The system can be figured such that all or a part of the atomic instructions reference the same control bit and status bit or different control and status bits may be specified for different types of instructions requiring different amounts of time for execution and therefore varying the amount of the delay 95. Alternately, the delay 95 can be programmable according to the instruction to be used.
Another example of an atomic instruction executed in accordance with the architecture of the present invention is shown in FIG. 3b. FIG. 3b shows a simple move register instruction which, for purposes of illustration, is identified as an atomic instruction. Immediately following the move instruction, a set bit instruction is executed to set the control bit. A branch if clear statement is then executed which monitors the status bit to determine when the status bit is set, thereby permitting execution of subsequent instructions.
The invention has been described in conjunction with the preferred embodiment. It is evident that numerous alternatives, modifications, variations, and uses will be apparent to these skilled in the art in light of the foregoing description.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4218739 *||Oct 28, 1976||Aug 19, 1980||Honeywell Information Systems Inc.||Data processing interrupt apparatus having selective suppression control|
|US4503495 *||Jan 15, 1982||Mar 5, 1985||Honeywell Information Systems Inc.||Data processing system common bus utilization detection logic|
|US5148544 *||May 17, 1991||Sep 15, 1992||Digital Equipment Corporation||Apparatus and method for control of asynchronous program interrupt events in a data processing system|
|US5175829 *||Oct 25, 1988||Dec 29, 1992||Hewlett-Packard Company||Method and apparatus for bus lock during atomic computer operations|
|US5218712 *||Jul 11, 1990||Jun 8, 1993||Digital Equipment Corporation||Providing a data processor with a user-mode accessible mode of operations in which the processor performs processing operations without interruption|
|US5274823 *||Mar 31, 1992||Dec 28, 1993||International Business Machines Corporation||Interrupt handling serialization for process level programming|
|US5363506 *||Oct 19, 1993||Nov 8, 1994||Kabushiki Kaisha Toshiba||Information processor with delayed interrupt device|
|1||*||Intel i960 CA Microprocessor User s Manual, pp. 12 15, Intel 1992.|
|2||Intel i960™ CA Microprocessor User's Manual, pp. 12-15, Intel 1992.|
|3||*||Sowell, Programming in Assembly Language: Macro 11, 1984, Addison Wesley Pub. Co. pp. 174 176, 313 326.|
|4||Sowell, Programming in Assembly Language: Macro-11, ©1984, Addison-Wesley Pub. Co. pp. 174-176, 313-326.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US6128710 *||Jun 5, 1998||Oct 3, 2000||International Business Machines Corporation||Method utilizing a set of blocking-symbol resource-manipulation instructions for protecting the integrity of data in noncontiguous data objects of resources in a shared memory of a multiple processor computer system|
|US7181600||Aug 2, 2001||Feb 20, 2007||Mips Technologies, Inc.||Read-only access to CPO registers|
|US7185183||Aug 2, 2001||Feb 27, 2007||Mips Technologies, Inc.||Atomic update of CPO state|
|US7502917 *||Dec 5, 2002||Mar 10, 2009||International Business Machines Corporation||High speed memory cloning facility via a lockless multiprocessor mechanism|
|US7539818 *||Jul 31, 2006||May 26, 2009||Blue Coat Systems, Inc.||Network object cache engine|
|US7600100||Dec 6, 2006||Oct 6, 2009||Mips Technologies, Inc.||Instruction encoding for system register bit set and clear|
|US7634638 *||Oct 22, 2002||Dec 15, 2009||Mips Technologies, Inc.||Instruction encoding for system register bit set and clear|
|US8788794||Dec 7, 2010||Jul 22, 2014||Advanced Micro Devices, Inc.||Programmable atomic memory using stored atomic procedures|
|US8793471||Dec 7, 2010||Jul 29, 2014||Advanced Micro Devices, Inc.||Atomic program verification|
|US9122476||Dec 7, 2010||Sep 1, 2015||Advanced Micro Devices, Inc.||Programmable atomic memory using hardware validation agent|
|US9760511 *||Oct 8, 2014||Sep 12, 2017||International Business Machines Corporation||Efficient interruption routing for a multithreaded processor|
|US20040111547 *||Dec 5, 2002||Jun 10, 2004||International Business Machines Corporation||High speed memory cloning facility via a lockless multiprocessor mechanism|
|US20070050662 *||Jul 31, 2006||Mar 1, 2007||Michael Malcolm||Network Object Cache Engine|
|US20070234020 *||Dec 6, 2006||Oct 4, 2007||Mips Technologies, Inc.||Instruction encoding for system register bit set and clear|
|US20100031005 *||Oct 9, 2009||Feb 4, 2010||Mips Technologies, Inc.||Instruction Encoding For System Register Bit Set And Clear|
|US20150074311 *||Oct 31, 2013||Mar 12, 2015||International Business Machines Corporation||Signal interrupts in a transactional memory system|
|US20160103774 *||Oct 8, 2014||Apr 14, 2016||International Business Machines Corporation||Efficient interruption routing for a multithreaded processor|
|U.S. Classification||712/203, 710/267, 712/E09.079, 712/E09.063, 710/262, 711/155, 712/E09.049|
|International Classification||G06F9/32, G06F9/38|
|Cooperative Classification||G06F9/30094, G06F9/3004, G06F9/30087|
|European Classification||G06F9/30B, G06F9/30A2, G06F9/30A8S|
|Feb 26, 1993||AS||Assignment|
Owner name: INTEL CORPORATION, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNOR:GANDHI, JAYANTI L.;REEL/FRAME:006447/0988
Effective date: 19930222
|Jun 22, 2001||FPAY||Fee payment|
Year of fee payment: 4
|Jun 23, 2005||FPAY||Fee payment|
Year of fee payment: 8
|Jun 17, 2009||FPAY||Fee payment|
Year of fee payment: 12